Disease tests: Risky business

At first glance, a new screening method for cardiovascular disease seemed like a major advance.

According to a study published last month in The New England Journal of Medicine, healthy adults who flunked a panel of 10 heart-related lab tests were four times likelier to die over a period of about 7 1/2 years than patients with good scores, even after controlling for factors already linked to heart disease like high cholesterol.

But surprisingly, the researchers -- most of whom are based in Boston -- concluded that their screening method wouldn't be very good at predicting heart problems in healthy patients.

The trouble, said James Ware, a statistician with the New England Journal who wrote an accompanying editorial, is that it's much easier to identify patterns among a group of people than it is to predict risk in individuals.

The study serves as a cautionary tale to many other scientists and companies attempting to develop new screening tests using so-called biomarkers -- substances in the blood or urine linked to diseases like cancer and heart attacks. Abnormal levels may be related to one step of a disease process, but often a whole chain of events must occur before a person falls ill.

In this case, the researchers measured substances in the blood and urine loosely associated with heart disease and hoped to show that all 10 tests taken together could help doctors better predict risk. But although lab abnormalities were considerably more common among patients who developed heart disease or died during the study, the tests were wrong far too often to be of much use.

"If you had a patient in your office with highly abnormal lab tests, it would be difficult to say much about specific risk," said Dr. Thomas Wang, a cardiologist at Massachusetts General Hospital and the study's lead author. "The patients we tested were healthy, and most didn't develop any problems over five to 10 years regardless of their test results."

According to Ware, even well established risk factors for heart disease -- such as high cholesterol, smoking, diabetes, and high blood pressure -- are of little use for predicting risk. In the New England Journal study, these risk factors were markedly abnormal in only about 30 percent of people who ended up dying.

"Screening tests must be quite a bit more accurate than you might initially think to correctly identify people who will develop a disease," Ware said.

In contrast, diagnostic tests for acute illness do not need to achieve the same degree of accuracy.

"When a patient is having chest pain, a positive test for a heart attack" means that the patient "is most likely really having a heart attack," Ware said. Unlike healthy people getting a screening test, patients with chest pain are sick to begin with, so it is much easier to believe a positive test result, he said.

Another shortcoming of screening tests is that even the best ones often fail to pick up many people who are at risk. This is particularly true for common ailments like heart disease that result from several factors such as genetics and environmental exposures. No one factor -- or even 10 -- can capture all the risk.

Last month's study highlights this point as well.

"Most people who developed a problem had a normal test result," Wang said. "In other words, the test wasn't very sensitive."

But the challenges don't stop there: There's also the question of whether doctors can treat the problems they find and whether the treatments are effective.

Few doubt, for instance, that screening older men with a blood test called PSA can help catch prostate cancer early.

What is less clear is whether patients with prostate cancer benefit from early treatment, and no study has demonstrated this convincingly. Because the treatments are invasive and cause long-term side effects and because many prostate tumors grow slowly, some experts believe men with early stage disease may be better off without any therapy at all.

Similarly, in a study published earlier this month in the Journal of the American Medical Association, researchers from California showed that blood levels of a protein called NT-proBNP were effective at predicting whether patients with known cardiac problems would have a heart attack. (Since all subjects in the study had heart disease to begin with, the task of identifying an effective screening test was made much easier.)

But NT-proBNP testing is by no means ready for clinical use.

"After you show that you can identify a disease early, you're left with the, 'So what'?" said Dr. Marvin Konstam, a cardiologist at Tufts.

It's not yet possible to lower NT-proBNP levels in high-risk patients, and even if it does become possible to do so, no one knows whether reducing the levels will help patients avoid a heart attack.

Even though such disease markers have yet to prove their usefulness for screening healthy patients, however, they are helping basic scientists learn more about the underlying mechanisms of disease.

"No one knows why molecules such as NT-proBNP consistently seem to show up" in abnormal levels in patients with heart disease, Konstam said. "If we can figure it out, maybe it will help us develop more effective treatments."

And experts are optimistic that more effective tests will come along eventually.

"Researchers are now doing systematic surveys of people who have a disease and those who don't," Wang said. "This should help us identify novel disease markers that no one would have thought to look at before."